Welcome to TechFreedom Dispatch #7
Here in the northern hemisphere it’s definitely autumn. As we approach Halloween, it’s the perfect time to be thinking about how ‘spooky’ Big Tech companies can be, and to consider your options… 🦇🎃💀
Quick updates:
- Our ‘shared inference’ session went well last month, and Tom’s busy working away on our fork of SwarmLLM. More on that next time, as we’ve been invited to the University of Bristol for a ‘Just Enough AI’ event in conjunction with Friends of the Earth.
- It was a shame that we didn’t get enough sign-ups to make viable the TechFreedom cohort we’d scheduled for this month. If you’d considered joining us, but decided not to, it would be really helpful to discover why. Whether it’s cost, timing, or something else, please tell us.
- The CoTech gathering last month was great, and we’re looking forward to heading to the Mozilla Festival in Barcelona in a couple of weeks’ time. If you’re going to be there too, let us know!
One other thing, we received a small amount of funding from Spark Grants to run a Digital Sovereignty Unconference in Newcastle-upon-Tyne in early 2027. If you’re anywhere nearby and would like to receive updates, sign up here!
In this dispatch: friction, opt-out October, radically open AI models, digital sovereignty, and owning the stack.
Read on!
Our stuff
Finding friction
Tom uses a climbing analogy to make a great point about how, in a world of easy answers, what we really need is space and time to reflect, adapt, and evolve. What we need is productive friction. Also check out his follow up post about slack (no, not the Salesforce-owned platform!)
Link
A European voice assistant for the Rabbit r1
Doug has created Carrot, a front-end using a European voice assistant for his Rabbit r1. Carrot uses models provided by Mistral and is designed to be intuitive and private, with features such as reminders, to-do lists, and weather updates. The project aims to provide a TechFreedom-compliant alternative to traditional voice assistants, with a focus on data minimisation and user control.
Link
Things that have caught our attention
1. Welcome to Opt Out October. Let’s Take Control of Our Data and Our Devices
The Electronic Frontier Foundation (EFF) has dubbed this month ‘Opt Out October’, arguing that "Big tech companies are hell-bent on taking away our agency and forcing us to use their surveillance-based services, apps, and devices."
We "don’t have to stand in the middle of the information super highway and let companies run us over," however. Whereas last year, they shared tips on how to reduce the amount of data companies harvest from you, this year "we’re looking to abandon those gears altogether and cut that data collection off at the head".
Radical. As they say, Big Tech companies claim that it’s somehow inevitable that you’ll use their platforms. And, while sometimes using a Big Tech platform might be the best decision for you and your org, it shouldn’t be the default . The easiest way to find alternatives is to start looking for them.
Link
2. Paying for frontier AI models buys 4-month head start at 5x the cost
Ars Technica reports that, according to The State of Open Source AI report by Mozilla, "many companies are shifting to the significantly cheaper open models for routine work" and that "most organizations should ideally be using open models as the default for the majority of their work".
For example, despite all of the publicity given to Anthropic’s Fable 5 model, Moonshoot AI’s Kimi K3 "achieves a composite AI performance score… that is just three points behind… all while costing just 30 percent of the latter."
As Raffi Krikorian, Mozilla’s CTO commented, "We see the decision to pay for closed [models] as workload-specific rather than organization-specific." Now is the time to be thinking about how, where, and why you’re using open (and in some cases local models) within your organisation
Link
Talking of open models, how about one that’s radically open? The Institute of Foundation Models (IFM) has released K2 Horizon, "a connected fleet of six models" which, "cross reasoning, mathematics, coding, agentic tasks, and general capabilities… delivers top-tier performance in every size class".
Why ‘radically’ open? AI models are usually like getting a finished loaf of bread, while what the IFM has released provides the recipe. This extends beyond initial training into "post-training" which is a name given to processes used to develop instruction-following, reasoning, coding, planning and tool use.
This might be a good one to share with people who throw their hands in the air and just accept that we must accept whatever Big Tech throw at us. It’s still early days for this level of openness, but we’re here for it! See also open-1b.
Link
4. Governance Analysis: How is Digital Sovereignty Measured Beyond Physical Location?
We mentioned in the introduction that we’re running a Digital Sovereignty Unconference. But what do we even mean by the term ‘digital sovereignty’?
In an article Bhavyatta Bhardwaj explains TechFreedom lens one (jurisdiction) in a Canadian context. They explain the difference between data residency , which asks simply "where’s the server?" , and digital sovereignty which asks a whole slew of questions, such as:
-
Who holds the encryption keys?
-
Who has administrator access?
-
Which company ultimately controls the system?
-
Which country’s laws apply?
-
Where do backups, logs and derived data go?
-
Who can keep the service running (or switch it off) during a dispute or emergency?
Bhardwaj introduces what they call the Lovelace Threshold as the point at which a policy claims to govern something but cannot see or control the underlying conditions that determine the outcome.
To continue the Canadian context, it’s a bit like renting a locked room in Ontario, but the landlord is in another country, keeps the master key and decides who may enter. While the room is physically in Canada, meaningful control still sits elsewhere.
Link
5. Government turns to open source for digital independence
We’ve discussed the case of the US, Microsoft, and the ICC both in our dispatches and in the pilot TechFreedom cohort. Now, European governments are doing something about it.
"The Dutch central government, provinces, municipalities and educational institutions will increasingly use open-source code in their software… The move is part of a broader debate about the government’s digital dependence. The choice to use open source therefore goes beyond how software is developed. The agreement is intended to give the government greater control over the digital infrastructure on which public services rely and make it less dependent on individual software suppliers."
We’ll be keeping an eye on this movement as it spreads.
Link
Other links
Things that we came across that almost made the cut for a write-up
That’s all for now! Thanks for reading.
Doug & Tom :)
PS If you know of someone who would value this dispatch, why not forward it to them?
TechFreedom is an initiative of FIELD STATION